The sensitivity of poverty analysis to dimensionality and distribution sensitivity: Evidence from district level of Pakistan

An awareness of how the findings from poverty analyses may change following introduction of multidimensional poverty measures is crucial for policy makers. This thesis highlights the changes in results of poverty analysis in Pakistan, which is a setting where multidimensional poverty measures recently either supplanted or supplemented money-metric poverty measures. The analysis relies on district-level poverty estimates, every second year from 2004 to 2014, calculated from the Pakistan Social and Living Standards Measurement survey and Household Income and Expenditure survey. Around two-thirds of districts show an opposite movement in poverty trends when comparing money-metric and multidimensional poverty trends. The convergence in poverty rates across districts is evident for money-metric poverty measures but not for multidimensional poverty measures. Relatedly, spillover effects on nearby districts matter for convergence in money-metric poverty but not for multidimensional poverty. The districts with high initial money-metric poverty estimates are catching-up to other districts but the districts with high initial multidimensional poverty are not. For reducing money-metric rural poverty (in particular the Squared Poverty Gap index) the growth of secondary towns is significant; however, for reducing multidimensional rural poverty it is not. The unconditional cash transfer programme, Benazir Income Support Programme (BISP), helps to reduce district-level money-metric poverty more than multidimensional poverty. Community-based targeting used in identification of poor households under BISP is more effective than Proxy Means Testing for money-metric poverty but the opposite is true for multidimensional poverty. Hence, given these varied patterns for money-metric and multidimensional poverty, policy makers need to be cautious when they draw conclusions from poverty analysis, particularly in settings where the multidimensional measures are either supplementing or supplanting money-metric poverty measures relied on previously